Sie sind auf Seite 1von 26

# The absolute beginner’s guide to coding Bytebeats!

## The absolute beginner’s guide to coding Bytebeats!

v1.1, 2018-09-18

written by
The Tuesday Night Machines
hello@nightmachines.tv

1 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Good morning!

Thank you for checking out my introduction to Bytebeat coding! When I discovered
Bytebeats, I was amazed how a few characters of text could generate such a vast variety of
awesome lo-fi sounds.

I’m not a good mathematician or programmer though, so at first I believed that I would never
be able to code a Beat without resorting entirely to trial and error. Of course randomly
hacking Bytebeats was still fun, but after a while, I really wanted to know why things
sounded the way they did and have the the ability to deterministically add elements to my
Beats. Unfortunately, specific information for absolute beginners was surprisingly hard to find
online, so after a bunch of long train-rides and revisiting old math and programming lessons,
I finally had some of the awaited epiphanies, resulting in a compilation of notes, which were
the basis for the document at hand.

This guide is for electronic musicians with no, or barely any, programming experience, who
enjoy experimenting with weird audio. You should know a little bit about waveforms, like
sawtooth, square and triangle waves, frequencies and amplitudes in an electronic music
context, as well as basic arithmetic: addition, subtraction, multiplication and division.

Anything else, I try to explain in an easy to follow manner, with fun exercises in between.
Let me know how this guide worked for you.

## Felix / The Tuesday Night Machines

2 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## What are Bytebeats? 4

Exercise 01: Experimentation 4

## What makes an expression a “Bytebeat”? 7

Exercise 02: Operators 7

## Coding with mathematical operators 11

Exercise 03: Multiplication and Division 11
Exercise 04: Modulo % 13
Exercise 05: Addition and Subtraction 13

## Coding with bitwise operators 14

Exercise 06: AND & 15
Exercise 07: OR | 16
Exercise 08: AND &, OR |, XOR ^ 17
Exercise 09: Bitshifts << >> 18

## Coding with relational operators 19

Exercise 10: Relational Operators 19

## Putting it all together 20

Coding a Step Sequencer 20
Tip: Line Breaks 22
Coding a Square Wave with PWM 23
Coding a Triangle Wave with Wave Shaping 23

## Sources & further reading 26

Version history 27

3 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## What are Bytebeats?

Bytebeats are musical creations from only a few lines of code, typically directed to an
unsigned 8 bit, 8 kHz audio stream. This means that a mathematical expression is
processed by the computer 8,000 times per second, resulting in an audible waveform with a
256-step resolution from silence (0) to full amplitude or “volume” (255).

For example:
(t*4|t|t>>3&t+t/4&t*12|t*8>>10|t/20&t+140)&t>>4

Oh man … that looks super abstract and complicated! How is that fun?!

Honestly, when I wrote it, I had almost no clue what was going on. The above example was
the result of pure trial and error and I assume that many people code Bytebeats purely in this
fashion. It is still lots of fun, because it’s quick and easy to do and happy accidents are
plenty. It’s the ultimate weird sample factory for your daily commute :D

## Exercise 01: Experimentation

Open Greggman’s HTML 5 Bytebeat Player using ​this link​ and press play. The variable ​t
generates a sawtooth wave. Leave this as it is. Now copy and paste some of the following
lines of code after ​t​ and see what happens.

|t*4
|t%128*10
|t*5^t/30
|t/20
&120
^t*7

Feel free to rewrite the code too, by changing values and operators! Notice how the lines
above all start with a logical operator like ​OR ​|​, ​AND ​&​ and ​XOR ​^​. This is required to
combine several statements after the initial ​t​.

If you encounter interesting sounds or melodies, save the code in a text file, Google Doc
or write them in your diary.

I hope that was somewhat enjoyable! Who needs programming knowledge anyway?

However, in this introduction I would still like to teach you some of the basics of Bytebeat
coding, so that you can combine the trial and error results with more predictable elements.

4 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## How to code Bytebeats?

Various applications exist to make coding, playing and recording Bytebeats more accessible,
compared to ​doing it manually in the Linux Terminal​. Here are three examples.

Damien Quartz’ Evaluator​ is a full-featured Bytebeat player with MIDI support for macOS
and Windows. It's available as a standalone app or VST plugin:
https://damikyu.itch.io/evaluator

Kymatica’s BitWiz iOS App​ offers a clean, multi-line coding interface, an oscilloscope,
performance controls and inter-app audio:
http://kymatica.com/Software/BitWiz

Single Cell’s Caustic 3​ for Android, iOS, Windows and macOS is a full-featured DAW
including a Bytebeat synthesizer, called “8BitSynth”:
http://www.singlecellsoftware.com/caustic

5 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Earlier, I wrote that a Bytebeat is a mathematical expression, executed through an 8 bit, 8

kHz audio output, so one might think that one can just write anything imaginable with enough
knowledge of math and programming. But what makes the result a “Bytebeat” then exactly?

Bytebeats can be considered a form of ​low-complexity art​, with ties to the ​Demoscene​,
where the aim is to create art in often incredibly limited environments.

## For Bytebeats, these limitations have typically been:

● One expression only
● Only one pre-defined variable: ​t
● Mathematical operators: ​( ) + - * / %
● Bitwise operators: ​& | ^ << >>
● Relational operators: ​< > <= >= == !=

Of course, those are just arbitrary limitations and you may do whatever you like. If you don’t
understand any of the above things yet, don’t worry. We’ll start learning about them right
away.

But first, try out the above mathematical, bitwise and relational operators in combination with
t​ and randomly selected numbers again. Just like we did earlier. See what you come up with
and save your cool code snippets!

## Exercise 02: Operators

Here is a ​new link​ to the HTML 5 Bytebeat Player with ​t/20​ as a starting point. Now, add
one or more of the following lines of code and experiment with replacing the operators with
the ones above.

|t*4*(t%10000>2000)&t>>4
|t&64
|t*t%128*(t%15000<2000)
|t

After clicking the previous exercises’ links to the HTML 5 Player, notice how the first lines
start with ​//​. This double slash denotes a “comment”, which is ignored by the compiler.
So any characters in a line of code after ​//​ won’t affect the result of our Bytebeat, which
means you can use ​//​ to take out code snippets, without actually having to delete them.
explaining what a piece of code does, so that you keep an overview if things get more
complex.

6 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Alright, after messing around, let’s see if we can actually figure out why a certain piece of
code behaves like it does. Open one of the above-mentioned editors and let’s start coding!
The simplest, audible Bytebeat, as you already know, is the following:

t

## It’s a sawtooth wave! Yay! … but why?!

The pre-defined variable ​t​ returns the time our Bytebeat is running. It is a timer, or counter,
which starts at zero on initial playback and increases by 1 on every new audio frame. So at
an 8 kHz sample rate, ​t​ increases by 8,000 every second and it just keeps going and
growing. This alone doesn’t create a sawtooth wave though, only a constantly rising number.
To make it a sawtooth wave, we need to have a look at our 8 bit audio output and some
binary basics.

“8 bit binary” means that there are eight digits, or bits, which can either be 0 or 1 and which
together can represent numbers from 0 to 255.

Each of the eight bits represents one specific number, from left to right:
128, 64, 32, 16, 8, 4, 2, 1

The numbers corresponding to the bits which are set to 1, are added together to create any
number from 0 to 255.

## 00000000 = 0 ​(all bits 0)

00000001 = 1 ​(first bit = 1)
00000010 = 2 ​(second bit = 2)
00000011 = 3 ​(first and second bits = 1+2)
00000100 = 4 ​(third bit = 4)
...
00010000 = 16 ​(fifth bit = 16)
00010001 = 17 ​(first and fifth bit = 1 + 16)
...
10000000 = 128 ​(eighth bit = 128)
10000001 = 129 (​ first and eighth bit = 1 + 128)
...
10001011 = 139 ​(first, second, fourth and eighth bit = 1 + 2 + 8 + 128)
...
11111111 = 255 ​(all eight bits = 1 + 2 + 4 + 8 + 16 + 32 + 64 + 128)

7 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Remember that ​t​ is growing by 8,000 each second? That’s way beyond our 0 to 255 range!
So what happens after the first split second when ​t​ becomes greater than 255? Obviously,
we can use greater numbers than 255 ​inside​ of our code, otherwise the whole ​t​ counter
thing wouldn’t work. It is only the ​output​ of our expression which gets truncated after 8 bits.
Looking at this process in binary again should make it clear how the sawtooth wave is
created.

Let’s say, for example, we have 10 bits available, letting us count from 0 to 1023. So we
have 10 digits, each 0 or 1, representing the following numbers:
512, 256, 128, 64, 32, 16, 8, 4, 2, 1

If we count up from zero, the 10 bits will fill up with 1 from 0000000000 to 1111111111, from
right to left:

0000000000 = 0
000000000​1​ = 1
00000000​10​ = 2
00000000​11​ = 3
0000000​100​ = 4
0000000​101​ = 5
0000000​110​ = 6
0000000​111​ = 7
000000​1000​ = 8
000000​1001​ = 9
000000​1010​ = 10
000000​1011​ = 11
000000​1100​ = 12
000000​1101​ = 13
000000​1110​ = 14
000000​1111​ = 15
00000​10000​ = 16
00000​10001​ = 17
00000​10010​ = 18
00000​10011​ = 19
00000​10100​ = 20
...
00​11111110​ = 244
00​11111111​ = 255

## This is also visualized in ​this video​.

8 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

As we said, we are using 10 bits for this example, so we have 10 digits available and can
count further up to 1023. When our ​output​ is truncated after 8 bits though, see what
happens when we count past 255, which is the 8 bit maximum value.

## 0011111111 = 255 (​ 255 in 10 bit)

00​11111111 = 255 (​ 255 truncated after 8th bit, nothing changes)

## 0100000000 = 256 (​ 256 in 10 bit)

01​00000000 = 0 (​ 256 truncated after 8th bit)

## 0100000001 = 257 (​ 257 in 10 bit)

01​00000001 = 1 (​ 257 truncated after 8th bit)

## 0100000010 = 258 (​ 258 in 10 bit)

01​00000010 = 2 (​ 258 truncated after 8th bit)

So our counter ​t​ can grow way beyond 255 internally in our expression, but the expression’s
result, processed through the 8 bit output, wraps around to 0 after 255, counts up to 255,
then falls down to 0 again, counts up to 255, back to 0 and so on. It’s an 8 bit, rising
sawtooth wave!

Let’s look at the chain of events inside our Bytebeat player once more, using the simple
expression ​t*50​:

## INPUT -> EXPRESSION -> RESULT -> OUTPUT TO 8 BIT

t=0 ​t*50​ 0 0 (​00​00000000)
t=1 ​t*50​ 50 50 (​00​00110010)
t=2 ​t*50​ 100 100 (​00​01100100)
t=3 ​t*50​ 150 150 (​00​10010110)
t=4 ​t*50​ 200 200 (​00​11001000)
t=5 ​t*50​ 250 250 (​00​11111010)
t=6 ​t*50​ 300 44 (​01​00101100)
t=7 ​t*50​ 350 94 (​01​01011110)
t=8 ​t*50​ 400 144 (​01​10010000)
...

How far can ​t​ grow? Up to 2,147,483,647, which is the maximum number when having 32
bits available, which seems to be the norm for Bytebeat players. At 8 kHz, i.e. 8,000 samples
per second, this would mean that it can count up for about 74 hours. What happens then? I
don’t know … it either starts back from 0, crashes, or it opens up a portal to the Chiptune
Dimension. Try it out :-P

9 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Coding with mathematical operators

Now, let’s modify our sawtooth wave with the available mathematical operators:
* / + - %

Multiplying​ ​t​ with a number makes ​t​ grow faster, increasing the sawtooth’s frequency.
For example:
t*2
t*4
t*12

Dividing​ ​t​ by a number makes ​t​ grow slower, decreasing the sawtooth’s frequency. This
can go way below audio rate, into super slow LFO territory.
For example:
t/2
t/4
t/30

Adding or subtracting​ numbers will apply an offset to ​t​, which by itself won’t do anything
sonically, because we’re just constantly adding/subtracting ​before​ the 8 bit output, which
wraps everything to a range of 0 to 255.

For example:
t+50
t+1000
t-300
Those lines do not have an audible effect on their own, but they do shift the phase, or
position in time, of the waveform ​t​, which becomes apparent when combining waves of
different phases. We will try this out shortly!

Another use-case of the minus operator is to ​invert​ a value, the same as if you’d multiply it
by -1.

## Exercise 03: Multiplication and Division

Open your Bytebeat editor of choice and enter any of the code snippets above, playing
with the multiplication and division of ​t​.

## Also try out:

-t
… the inversion of our sawtooth.

10 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Modulo​. Ah! Now things get interesting! The modulo ​%​ returns the remainder after a division.
What’s a remainder? Let’s do some simple math (that’s all I can do, in fact):

15 / 4 = 3.75

This means that 4 fits into 15 three whole times, plus a little bit. That little bit is the
remainder:

0.75 * 4 = 3

Or you could think of it this way, which I actually found easier to grasp at the beginning:

15 / 4 = 3.75

## Leaving away the 0.75, let’s only do:

3 * 4 = 12

… because we know that 4 fits into 15 three whole times. That makes 12, plus the
remainder, which is the difference between 15 and 12, i.e. 3.

## To get only the remainder 3, you would write the following:

15 % 4 = 3

That’s … uhm … cool, I guess?! But what does it mean for my sick Bytebeat coding? I’m
glad you asked! Let’s look at some more modulo calculations and see if we can spot a
pattern:

4 % 4 = 0
5 % 4 = 1
6 % 4 = 2
7 % 4 = 3
8 % 4 = 0
9 % 4 = 1
10 % 4 = 2
11 % 4 = 3
12 % 4 = 0

Notice something? That’s like a tiny sawtooth wave! The modulo of ​anything​ divided by 4, is
always 0, 1, 2 or 3! Or more generally speaking:

## x % y ​is always between ​0​ and ​y-1

So let’s code a tiny sawtooth Bytebeat and see what happens to our audio!

11 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Open your Bytebeat editor of choice and enter:

t%128

This will make the expression’s result something between 0 and 127, which means we
only use half of our 8 bit range from 0 to 255, reducing our sawtooth amplitude (or
“volume”) by half. Now try out different numbers instead of 128 and see what happens!

Well done! Have you noticed how the wave not only changes in amplitude, but also in pitch?
That means that you can also use the modulo to change the frequency of a waveform by
small amounts, not just large ones, like when you multiply by 2 or divide by 2 or more.

The modulo also acts just like our 8 bit output, i.e. it wraps the incoming values to a specific
range (like 0 to 127 above), only that we can now decide that range for ourselves inside our
expression. Of course our expression’s result is still truncated to 8 bits at the end.

A while back, I told you we’d revisit ​addition and subtraction​. Now is that time!

t%128

## Yawn! Okay, okay! Enter something new:

t%128+128

Now our sawtooth isn’t going from 0 to 127 anymore, but from 128 to 255! This doesn’t
change its amplitude (still 127 steps), but it changes the values we’re working with.

Try adding or subtracting other values and see what happens, especially when you offset
the sawtooth to go partly below 0 or above 255.

## Time for some phase shifting!

t%255+t%64
This adds two ​t​ waves of different frequencies and amplitudes together, resulting in a
sound similar to the well known “Phaser” effect.

12 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Coding with bitwise operators

Working with mathematical operators was probably not much news to you, so let’s try out our
available bitwise operators, which perform tasks based on a number’s binary representation.
& | ^ << >>

AND &​ compares the bit values of two binary numbers and returns 1 only if ​both​ bits are 1.

010​1​ = 5
& 001​1​ = 3
= 000​1​ = 1 = 5&3

10​11​ = 11
& 00​11​ = 3
= 00​11​ = 3 = 11&3

0011 = 3
& 0100 = 4
= 0000 = 0 = 3&4

Yes, ​AND ​&​ only outputs 1 if both bits are 1. It’s like ​multiplying​ both bits:
0 * 0 = 0
1 * 0 = 0
1 * 1 = 1

## Now spot the pattern in those examples:

11​1​1111100 = 1020
& 00​1​0000000 = 128
= 00​1​0000000 = 128

1100001110 = 782
& 0010000000 = 128
= 0000000000 = 0

10​1​1011101 = 733
& 00​1​0000000 = 128
= 00​1​0000000 = 128

1001011010 = 602
& 0010000000 = 128
= 0000000000 = 0

Looks like if we write something like ​t&128​, we’ll only get two alternating values: 128 and 0.
What does that look and sound like?

13 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Open your Bytebeat editor and enter:

t&128
It’s a square wave! Replace the 128 with other values and see what happens.

Next, combine our sawtooth ​t​ with a much slower version of itself, enter:
t&t/30
That’s some sweet ducking!

What happens when you invert that slow sawtooth wave using the minus operator in front
of it?
t&-t/30

A curious property of an ​AND ​&​ operation is that the result cannot be larger than the
smallest operand. For example:

t&140
… can never grow past 140, no matter how high ​t​ gets.

OR |​ compares the bit values of two binary numbers and returns 1 if ​one or both​ bits are 1.
This operator is thusly called “inclusive OR” or sometimes “and/or”.

0​101​ = 5
| 0​011​ = 3
= 0​111​ = 8 = 5|3

​1​0​11​ = 11
| ​0​0​11​ = 3
= ​1​0​11​ = 11 = 11|3

0​011​ = 3
| 0​100​ = 4
= 0​111​ = 8 = 3|4

So ​OR ​|​ just takes any 1 it gets and keeps it. It’s like ​adding​ both bits together:

0 + 0 = 0
1 + 0 = 1
1 + 1 = uhm ...

Okay, sorry, not entirely like ​adding​ the bits together ... I just wanted to make the point that
you can use ​OR ​|​ to ​add​ waveforms together, almost like with a super lo-fi audio mixer.

14 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Exercise 07: OR |

## Open your Bytebeat editor of choice and enter:

t
|t*4
|t*12

Use ​OR ​|​ to add more sawtooth waves of different frequencies (​t​ multiplied or divided by
a number).

A property of ​OR ​|​ that makes it behave so similar to an audio mixer is that the results can
never be smaller than the smallest operand (or “audio input” in our mixer analogy). So:

t|140
… will always result in a number of 140 or greater, no matter how low ​t​ is.

XOR ^​ is an “exclusive OR” which returns 1 only on non-matching bits, i.e. if they are 1 and
0:

0​10​1 = 5
^ 0​01​1 = 3
= 0​11​0 = 6 = 5^3

​1​011 = 11
^ 0​ ​011 = 3
= 1 ​ ​000 = 8 = 10^3

0​01​1 = 3
^ 0​10​0 = 4
= 0​11​0 = 6 = 3^4

Unlike our other two operators above, the result of an ​XOR ​^​ operation is not limited to a
minimum or maximum number.

15 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Exercise 08: AND &, OR |, XOR ^

If you haven’t already tried it out by yourself, go ahead and mix some waveforms using all
of the above logical operators: ​AND ​&​, O
​ R ​|​ ​and ​XOR ​^​.

Here’s a fun example to get you started. First, try to understand what kind of waveform
each line of code generates (slow or fast? sawtooth or square?) and then listen to the
output when you combine them using different logical operators.

t&128
| t&64
& t/20
^ t/15
| t*2

As already stated, the ​OR ​|​ operator provides the most similar results to a normal audio
mixer. Although, since we’re talking about Bytebeats, “similar” is still quite far off. Audibly,
AND ​&​ works too in many cases, but not all, while ​XOR ​^​ often breaks up the signal
significantly. It really depends on your expression though, so never stop experimenting.

What?! You’re still here? Okay, let’s check out the remaining two bitwise operators then!

Bitshift left << and Bitshift right >>​ work a little different than the rest, as they don’t
compare two binary numbers. Instead the only look at the first operand in binary and shift its
bits left or right, by the second operand’s value.

12 << 3
0000​1100​ = 12
<---
0​1100​000 = 96

All bits are simply shifted 3 places to the left; new bits are 0. This equals the following:

12 * 2​3​ = 96
or
12 * 2 * 2 * 2 = 96

16 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Now the other way around, bitshift right:

12 >> 3
0000​1100​ = 12
--->
0000000​1​ = 1

Again, all bits are shifted 3 places, this time to the right. This equals the following:

12 / 2​3​ = 1
or
12 / (2 * 2 * 2) = 1

So why not simply use regular, mathematical multiplication and division operators? Bitshifts
can be processed faster by the computer, which might create different results, depending on
the complexity of your code … however, that fact probably never actually applies to the
rather simple Bytebeats. But there’s another neat thing about using bitshifts in our
Bytebeats, as we will find out now.

## Exercise 09: Bitshifts << >>

Shift those Bitz and pay attention to the audio. How does ​t​’s frequency change, when
shifting it by 1, 2, 3, etc.?

t<<1
t<<2
t<<3

t>>1
t>>2
t>>3

Shifting ​t​ by 1, makes its frequency go up or down one octave! Good to know, when you
want to code traditionally western, musical Bytebeats!

17 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Coding with relational operators

One more category to go! Relational operators find out whether a certain relation between
two numbers is “true” or “false” and output 1 or 0 accordingly:

## 2 > 1 = 1 ​(“two is ​greater than​ one is true”)

2 < 1 = 0 ​(“two is ​less great than​ one is false”)
2 == 1 = 0 ​(“two is ​equal to​ one is false”)
2 != 1 = 0 ​(“two is ​not equal to​ one is true”)
2 >= 2 = 1 ​(“two is ​greater than or equal to​ two is true”)
2 <= 2 = 0 ​(“two is ​greater than or equal to​ two is true”)

We can use relational operators to create on/off switches for parts of our code, which takes
our Bytebeats to another level!

## Exercise 10: Relational Operators

Stop your editor’s playback and make sure that ​t​ is reset to zero (in Greggman’s HTML 5
Bytebeat Player, click on the time value between the ? and the play button). Enter all of
the following code and press play.

(t>8000)*t
|(t>16000)*t*2
|(t>24000)*t*6
|(t>32000)*t*12
|(t>40000)*t*40
|(t>48000)*t/20

After six seconds you can press stop, reset ​t​ to zero, and start again, if you like.
Remember that the output of a relational operation is 1 (true) or 0 (false)? By multiplying
this output with part of our expression, we can switch that part on or off, because:
0 * x = 0
1 * x = x

In the above example, we get 8,000 samples of silence, which at 8 kHz is one second.
Then, when ​t > 8000​ ​is true, the result of the first line of code reads:
(1)*t

Our sawtooth wave starts playing! One second later, at ​t > 16000​, we get:
(1)*t
|(1)*t*2

… a higher pitched sawtooth wave is multiplied by 1, i.e. it is added added to our output
via the ​OR ​|​ operator. Now go ahead and experiment with the other relational operators
and as usual, save fun code snippets to your Bytebeat diary!

18 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Putting it all together

In this chapter, I would like to show you step-by-step examples of applying the various
concepts we’ve learned in one Bytebeat.

## Coding a Step Sequencer

Let’s build a traditional 16-step sequencer, which loops sixteen sounds indefinitely. For this,
we need a looping counter, indicating the different step positions in our sequence. We have
already worked with such a revolving counter. Can you remember?

## Look at this piece of code:

t%16

The ​modulo %​ operator! In this case, ​anything​ % 16 will result in values between 0 and 15,
i.e. 16 steps. How cool is that? Of course, at the 8 kHz sample rate at which ​t​ is counting
upwards, those 16 steps will be very short, so let’s make them longer:
t%16000

Now every one of our 16 steps can be 1,000 audio frames long, or 1/8th of a second, at 8
kHz. That should be enough.

## Next, we need to define the actual steps:

Step 1 should last from 0 to 999
Step 2 from 1,000 to 1,999
Step 3 from 2,000 to 2,999

Step 15 from 14,000 to 14,999
Step 16 from 15,000 to 15,999

## How could we possibly implement this? Exactly: relational operators.

(t%16000 >= 0 & t%16000 < 1000)

## The above code will output 1 (true) as long as …

… the step counter ​t%16000​ is g ​ reater than or equal to​ 0
AND
… the step counter is ​less than​ 1000

So from 0 to 999, this will output 1 (true). Of course, this by itself won’t make a sound, but
we can multiply the result (true or false, 1 or 0) with trusty old ​t​:
(t%16000 >= 0 & t%16000 < 1000) * t

Try the above line in your Bytebeat player and then, add a second line with ​OR ​|​ for step 2,
from 1000 to 1999, which could play a higher note, like ​t*3​:
| (t%16000 >= 1000 & t%16000 < 2000) * t*3

… and then add more lines, for the remaining 14 steps in our sequence.

19 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Look at that! You turn the page and new code appears! Isn’t that great? Here’s the complete
sequencer code for all 16 steps, playing a simple sawtooth melody:

## (t%16000 >= 0 & t%16000 < 1000) * t

|(t%16000 >= 1000 & t%16000 < 2000) * t*3
|(t%16000 >= 2000 & t%16000 < 3000) * t*5
|(t%16000 >= 3000 & t%16000 < 4000) * t*8
|(t%16000 >= 4000 & t%16000 < 5000) * t*2
|(t%16000 >= 5000 & t%16000 < 6000) * t
|(t%16000 >= 6000 & t%16000 < 7000) * t*12
|(t%16000 >= 7000 & t%16000 < 8000) * t*8
|(t%16000 >= 8000 & t%16000 < 9000) * t*3
|(t%16000 >= 9000 & t%16000 < 10000) * t
|(t%16000 >= 10000 & t%16000 < 11000) * t*4
|(t%16000 >= 11000 & t%16000 < 12000) * t*10
|(t%16000 >= 12000 & t%16000 < 13000) * t*7
|(t%16000 >= 13000 & t%16000 < 14000) * t*6
|(t%16000 >= 14000 & t%16000 < 15000) * t*5
|(t%16000 >= 15000 & t%16000 < 16000) * t*6

The ​t​ sawtooth sound is seriously getting on my nerves by now, so please, let’s change it
into a square wave! Lucky for us, that’s easily done. Put the sequencer code into

(
sequencer code
)
&128

Revisit the ​chapter on bitwise operators​, if you don’t remember why ​&128​ creates a square
wave.

You know, on second thought, it’s not just the sawtooth wave that I got tired of. That
repeating melody could use some variation too. How about we transpose it up one octave
every other run-through?

## ● code for the transposition

● a counter and relational operation that switches the transposition on and off

Can you come up with the required expression? A possible solution is on the next page.

20 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Here’s the new line of code we need:

<< (t%32000 <= 16000)

To transpose our sequencer output up by one full octave, we can simply bitshift it to the left
by 1. To switch the bitshifting on and off every other playthrough, we add a relational
operation, with a counter twice a high as our sequencer’s, i.e. ​t%32000​. Then, we ask if the
counter is equal to or greater than 16,000 and if true, we know we’re on a second
run-through and get 1 from the operation. In this case the result of our above expression

<< (1)
… bitshift left by 1 (true), which is a transposition up one octave.

To make it work properly, we have to insert that line of code between the sequencer and our
square wave conversion:

(
sequencer code
)
<< (t%32000 <= 16000)
&128

Alright, this has been a very deterministic approach to building our step sequencer program,
without any trial and error or randomly written elements. Go ahead and tweak the code to
your liking now. For example, you could change each of the 16 steps’ sounds from a
sawtooth wave to something completely wild and different, you could also change the length
for which each sound plays (right now it’s 1,000 audio frames), or you could add some code
around the sequencer to completely warp its output.

## Tip: Line Breaks

While one-liners look impressive, they often don’t provide a good overview of what’s
actually going on in the code. Using line breaks and ​//​ comments helps to find your way
easily and change parts quickly.

## Cool, but inefficient:

t*30&(t%3000<1500)*t/10|-t>>3*t/100

## Uncool, but splendidly efficient (it’s a joy!):

t*30 //super high pitched sawtooth
&
(t%3000<1500)*t/10 // rhythmically breaking up the above sawtooth
|
-t>>3*t/100 // bitshifting everything to hell

21 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Coding a Square Wave with PWM

Square waves sound even nicer with pulse-width modulation applied, a very common effect
found on most regular synthesizers. If you don’t know about PWM yet, check out ​this video​.

Let’s see how we can implement this effect in our Bytebeat. First, we need to write a square
wave expression that actually gives us control over the time that the wave is high and low.
Here’s a possible solution:
(t%100 >= 50)*255

There is a counter ​t%100​, counting from 0-99. We then check if its output is equal to or
greater than 50. If this is true, we get 1, which we multiply by the amplitude (or “volume”
level) we want our square wave to have, something between 0 (silence) and 255 (full
volume). The above code gives us a 50% pulse-width, full-volume square wave. To alter the
pulse-width, we have to change the ​50​ to something else, between 0 and 100. Try it!

## To get the pulse-width moving, we can write the following:

(t%100 >= ​t/200%100​)*255

Here, the ​50​ was exchanged for ​t/200%100​, another counter, counting from 0 to 99 again.
However this one is not counting at an audio-rate speed of ​t​, but way slower at ​t/200​. Play
around with this speed value and also decrease the counter’s maximum value below 100, to
only apply the pulse-width modulation across part of the wave’s duty cycle.

## Coding a Triangle Wave with Wave Shaping

So far, we’ve only worked with very “harsh” sounding waves. Let us now code a triangle
wave.

Think about how we could approach this. We know that ​t​ provides a rising sawtooth and ​-t
an inversion of that, i.e. a falling sawtooth. If we switched between ​t​ and ​-t​ after each
cycle, we’d get a triangle wave. How long is one cycle of our sawtooth waves? 256, because
after that, the output wraps around and starts a new sawtooth. So for starters, we need to
switch between the two waveforms every 256 audio frames. A expression for this could be:
t * (t%512 >= 256)
|
-t * (t%512 < 256)

As long as ​t%512​ is equal to or greater than 256, play ​t​ and while ​t%512​ is less than 256,
play ​-t​. This will alternate between the rising and falling sawtooths. Try it out and see what
happens!

It almost looks correct, but there is one thing that messes up our smooth triangle wave. What
is it and how can we fix it? Try to find the solution yourself, before swiping to the next page.

22 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

The solution:
t * (t%512 >= 256)
|
(-t-1)​ * (t%512 < 256)

Hmmmm ...

Before, on the previous page, both sawtooth waves would always start their cycle at 0. This
meant that the rising ​t​ wave would start 0, 1, 2, 3, 4 and so on, until 255, after which the
falling ​-t​ wave would take over, starting 0, 255, 254, 253, etc. This resulted in the super
short, but sonically very significant, drop down to 0, when both waves met at the top.
Subtracting 1 from ​-t​ achieves an offset between the waves, so that ​-t​ takes over when it
is at 255 and not at 0.

Next, let’s try some wave shaping, by putting parentheses around the triangle wave code
and then moving the whole thing upwards continuously.

(
t*4*(t*4%512 >= 256)
|
(-t*4-1)*(t*4%512 < 256)
)
+t/60

Outside of the parentheses, we added ​+t/60​ which shifts up the triangle wave very slowly
and continuously. As soon as the triangle wave wraps around from 255 back to 0, it is cut
up, resulting in harsher, pulsating sounds.

Also note how we multiplied all of the ​t​ waves inside the parentheses by 4, to increase the
triangle wave’s pitch to something more audible.

23 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

## Breaking the chains

Until now, we’ve stayed well inside the commonly accepted, yet arbitrarily defined,
constraints of Bytebeat coding​. Let’s take a quick peek over the fence!

Sometimes you might want to alter parts of your Bytebeat code while it is being executed
without typing on your keyboard, recompiling the code and restarting playback; for example
during a live performance. A simple way to do this is by using other variables, apart from ​t​.
Kymatica’s BitWiz iOS App​, for example, offers a multi-touch XY performance pad, as well
as the ability to route external MIDI signals to variables in your code. So you could set it up
that the variable ​a​ receives CC data from a connected MIDI keyboard’s modulation wheel.

t*a
... would then allow you to change the pitch of the ​t​ sawtooth wave by turning the wheel,
while your Bytebeat is running. Of course that is just the tip of the iceberg and you can
create much more complex, but still playable, Beats using variables (think back to the
relational operators or “on/off switches”).

Another limitation we could ignore is “one expression only”. In ​Single Cell’s Caustic
8BitSynth​, you can write two separate expressions, which run at the same time, letting you
blend between them and even automate this. Other players will allow you to write several
expressions into a single text field, separated by commas, with only the last one being sent
to the audio output. This allows you to declare variables and do other things outside of your
sound expression.

s​=6,
(t*​s​)*(t*​s​%512>256)
|
(-t*​s​-1)*(t*​s​%512<256)

That’s the triangle wave from the previous page, but instead of the value 4, we have the
variable ​s​, for “speed” or the wave’s frequency. First, ​s​ is declared as the number 6, after
which we have a comma and then the second and last expression which will be played
through the speakers. To change the pitch of the wave now, you only need to alter ​s​ and not
type the new value four times in various places in your code.

And finally, certain Bytebeat programs, like ​Greggman’s HTML 5 Player​, might accept more
operators than the basic ones we discussed in this guide, for example ​sin​, ​cos​, ​tan​,
sqrt​, etc.

24 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

That’s it for this guide. Thank you very much for checking it out! I hope you enjoyed learning
me know :-)

Of course there is more on this topic out there, so here’s a collection of websites you might
find interesting, some of which have also been linked earlier.

Content by Ville-Matias “viznut” Heikkila, who started the whole Bytebeat thing in 2011:
http://viznut.fi/texts-en/bytebeat_exploring_space.pdf
http://viznut.fi/texts-en/bytebeat_algorithmic_symphonies.html
http://viznut.fi/texts-en/bytebeat_deep_analysis.html
https://countercomplex.blogspot.com/2011/10/some-deep-analysis-of-one-line-music.html

## Article with a bunch of other links:

http://canonical.org/~kragen/bytebeat/

## Reddit post with lots of cool Bytebeat code snippets:

70bef

Bytebeat software:
http://coleingraham.com/2013/04/28/bytebeat-shell-script/
https://github.com/greggman/html5bytebeat
https://damikyu.itch.io/evaluator
http://kymatica.com/Software/BitWiz
http://www.singlecellsoftware.com/caustic

## Various Wikipedia articles:

https://en.wikipedia.org/wiki/Low-complexity_art
https://en.wikipedia.org/wiki/Demoscene
https://en.wikipedia.org/wiki/Binary_number
https://en.wikipedia.org/wiki/Bitwise_operations_in_C

## More from The Tuesday Night Machines:

Modular Synth Basics:
https://nightmachines.tv/modularbasics
My Synthesizer Wardrobe:
https://nightmachines.tv/ikea-pax
Like my free content? Here are a few ways to support TTNM:
https://nightmachines.tv/support

25 / 26
The absolute beginner’s guide to coding Bytebeats!
by The Tuesday Night Machines

Version history

2018-09-17 v1.0:
- initial release for comments on the AE Modular forum: ​http://forum.aemodular.com

2018-09-18 v1.1:
- spelling and mark-up corrections
- additional content and editing (thanks to forum users ​thetechnobear​ and ​careck)​